I think there are three scenarios which can prompt this:
- An incorrect copy of a question from a textbook/notes from a given class.
- The XY problem, as Asaf mentions in the comments.
- A question which springs independently as curiosity of the user.
Note that they are not mutually disjoint. I will adress your questions for each of the cases.
I think the most easily manageable is number 1. It usually is very easy to spot a typo, and so we only need to ask something like: "are you sure that is what you mean, and not (...)?" If the user is not responsive, then disengage. If the question makes sense with the typo, leave it be (it may become a difficult question/trivial question etc). If the question doesn't make sense, apply "standard procedure": vote to close as whatever best reason you see fit (it most likely will be unclear what is being asked, or lacking context). Furthermore, a typo will probably not lead to a lot of edits (there may be cases where a lot of typos may indeed lead to a lot of edits, but these are rare. I don't recall any instance of this happening).
Number 2 and 3 can be wild beasts. As soon as a "cat and mouse" question begins to surface (i.e., the user has edited/commented sometimes changing it) and it does not look like a typo, I think the best course of action is to try and verify if you are dealing with the XY problem. I think something as simple as commenting "Are you falling in the XY problem? If so, please try to make explicit the question that motivated you, so we can show better ways to get there." If the user acknowledges this, problem solved. If not, then you probably are in case 3 which is the worst in my opinion.
Number 3 has the most potential to be stressful. In this case, the user probably has not a good grasp of what he is asking and will forget "trivial" cases (or treat cases which are not trivial as "trivial" because they are not the kind of thing they expect). Here, I don't think there is a clear-cut procedure. Try and talk with the user, and emphasize that changing a question in such a way that potential answers could be invalidated, even if there are none, is bad. If he is not satisfied with the question he made due to the possible answers then make another, making sure that the question is the right one this time. If the talk is productive, good. But as soon as you realize the talk is not being (or will not be) productive, disengage. If you feel the user is being remarkably rude or disruptive, I think flagging for moderator attention is appropriate. If an edit war begins to happen (people rolling back the question to a previous state but OP not agreeing), flagging also seems appropriate.
Another problem with number 3 is that it may be the case that it is "clear" (also a problem to determine when something is "clear" or not) that OP meant something else, and thus the edit is expected, but people rushed to answer the question for whatever reason. However, this situation will probably be a "one-edit" kind of thing, not several ones. I'm only mentioning this problem for completeness, since I think it is relevant and is one instance where OP has less of the blame.